Citation

Paulik R, Zorn C, Wotherspoon L, Sturman J. Int. J. Disaster Risk Reduct. 2023; 94: e103826.

Copyright

(Copyright © 2023, Elsevier Publishing)

DOI

10.1016/j.ijdrr.2023.103826

PMID

unavailable

Abstract

Residential buildings a high proportion of any country's fixed asset wealth. Despite their vulnerability to flood damage, detailed exposure information on characteristics and monetary value is often unavailable for flood risk assessment. In this study we model physical and non-physical building characteristics for object-level replacement valuation in New Zealand flood hazard areas. Public data is used with geospatial and supervised learning models to estimate building characteristics. Replacement values were calculated from object-specific unit values based on useable floor area and physical and non-physical characteristics. Learning model evaluation showed higher prediction precision from a larger explanatory variable range while models learned on characteristics from all urban and rural areas transferred with higher precision. National flood hazard area exposure for 441,384 residential buildings was estimated at NZD $218 billion replacement value. Exposure comprises 282,395 houses valued at NZD $213 billion and 158,989 appurtenant buildings valued at NZD $5 billion. Spatio-temporal analyses demonstrated a high proportion of replacement value exposure occurred in five major urban areas. Building construction in flood hazard areas peaked between 1960 and 1980. Construction slowed in subsequent decades though total floor area and replacement value has continued to increase. The study demonstrates detailed information from object-level modelling is critical for understanding spatio-temporal drivers of exposure in flood risk assessments at national to local scales.

Language: en

Keywords

Economic value; Exposure; Flood risk assessment; Flooding; Learning models; Residential buildings